Prototyping Customer Service with Large Language Models

This paper presents a prototype aimed at enhancing the customer service workflow of Centiro Solutions, focusing on efficiently resolving incidents related to software products and services. Leveraging a Large Language Model (LLM), the prototype enhances information retrieval, problem-solving procedures, and communication alignment with company values. The design process involved an innovation sprint, a collaborative effort between industry and academia, employing methodologies like design-thinking and agile development. Evaluation results indicate several benefits, including cost reduction, improved customer service, and streamlined training for customer service technicians. The paper proposes four hypotheses for future research, including the integration of language models into workflows and the importance of user involvement in developing routines for utilizing language models effectively.